OMAE INVESTIGATING THE RESPONSE OF OFFSHORE FOUNDATIONS IN SOFT CLAY SOILS

Proceedings of OMAE st International Conference on Offshore Mechanics and Arctic Engineering June 3-8, Oslo, Norway OMAE-857 INVESTIGATING THE RESPONSE OF OFFSHORE FOUNDATIONS IN SOFT CLAY SOILS B.W. Byrne Department of Engineering Science The University of Oxford M.J. Cassidy Centre for Offshore Foundation Systems The University of Western Australia ABSTRACT A series of tests were conducted in a drum centrifuge with the aim of investigating the performance of typical offshore foundations on soft normally consolidated clay. The foundations consisted of spudcan footings and suction caissons. These types of foundations are being considered for use in various offshore applications including as foundations for mobile drilling rigs (jack-ups) and offshore wind turbines. A special loading device was designed so that combined loading could be applied to the footing. This device could apply the same ratio of horizontal to moment loading as that applied to the foundations of mobile drilling units. The main aim of the investigation was to compare how the performance changes as the foundation is varied. This is important when considering the use of a jack-up rig for a permanent facility, a concept that is increasingly being considered. In such a case there are concerns about the long-term suitability of the spudcan footing, with the amount of sustainable rotational fixity being of particular interest. A total of 6 experiments were carried out investigating areas that include a) comparing the vertical loading response in both compression and tension, b) using a fixed arm to apply predominantly horizontal loading, and c) using a hinged arm to apply a distinct ratio of horizontal to moment loading. Interestingly in the case of the spudcan footing considerable back-flow of the soil was observed during the installation phase. The combined load response of spudcans under these conditions is an area that has not been investigated thoroughly. Keywords: Offshore Foundations, Suction Caissons, Spudcans, Jack-up Rigs, Drum Centrifuge, Clay. INTRODUCTION Alternative shallow foundation solutions and the amount of moment fixity they provide is an area of particular interest in the offshore industry. For instance most jack-up rigs currently use circular conical footings known as spudcans as their foundations. Typically these jack-ups have been used for shortterm drilling, maintenance or construction operations. However, now they are being considered for deeper water and harsher environments necessitating a more critical assessment of their performance under increased loading. Furthermore, several jack-ups have been considered for permanent deployment at oil and gas fields. This means that stricter reliability criteria and harsher environmental conditions must be applied in the design process. It is usual for designers to assume that there is no moment fixity at the spudcan. However, this is an incorrect assumption and is the subject of much current research []. This incorrect assumption can lead to a misunderstanding of the rigs behaviour during extreme events (see Cassidy et al. [] for instance). An alternative approach to improve performance against the increased design criteria mentioned above is to investigate the use of alternative fondation solutions. These include suction installed skirted foundations or suction caissons. A suction caisson is a shallow foundation that is skirted around the periphery. The skirts are forced into the ground by applying a net over-pressure across the top of the footing by withdrawing water from within the caisson compartment (i.e. sucking it into the ground). The skirt of the caisson may lead to a greater moment and horizontal capacity than the spudcan. At present there has been no study devoted to comparing the two footings when subjected to the same loading conditions. Copyright by ASME

This study aims to address this by providing experimental evidence of the different responses. There have been, of course, many numerical studies investigating the effects of skirts on shallow foundations. The experiments described can be used to indicate areas where numerical studies agree as well as highlight areas where further effort is required. Although the main motivation for this study has been to compare shallow foundation options for jack-up units there are benefits for other applications. An example is the increased emphasis on the development of offshore wind farms in Europe []. This will involve constructing and installing large numbers of wind turbines, where the hub-height might be m above the sea floor and the blades might span a diameter of 7m. The nature of the loading leads to large moment loads but very small vertical loads on the foundation system (a different load condition to the jack-up, which has a substantial vertical self-weight). Obviously for some locations piling will be adequate, however, at other locations shallow foundations may be preferable (see Figure and also Byrne and Houlsby [3]). One structure under consideration is a monopod, where the moment and vertical loads will be imposed directly onto the foundation. An alternative system might be a tripod structure with three caisson foundations. The moment loads will be transferred to compression/tension loads applied at opposite foundations. In this case the tensile loading on the upwind foundation may be critical. NOMENCLATURE c v Coefficient of Consolidation (L /T) D Diameter (L) e Eccentricity Factor H Horizontal Load (F) h Installed Depth of Caisson ( L) H Pure Horizontal Capacity (F) L Skirt Depth (L) M Moment Load (F) M Pure Moment Capacity (FL) N c Bearing Capacity Factor Strip N c Bearing Capacity Factor Axisymmetric R Radius (L) s u Shear Strength ( F/L ) t Skirt Wall thickness ( L) u Horizontal Displacement (L) v Vertical Velocity (L/T) V, V Vertical Load, Effective Vertical Load (F) V Pure Vertical Capacity (F) w Vertical Displacement (L) β, β Parabolic Weighting Factors γ Effective Unit Weight of Soil (F/L 3 ) θ Rotation (degrees) EXPERIMENTAL SET-UP The tests described here were performed in the drum centrifuge located at the University of Western Australia and shown in Figure. This facility, installed in 998, has a.m diameter and a maximum acceleration level of 85g. The soil sample is contained in the outer channel, which is 3mm high (vertically) and has a mm radial depth. A central set of actuators provide vertical and radial motion which, combined with a hinged leg (described in detail below), allows a combination of vertical, horizontal and rotational motion to be applied to the footing. By using two concentrically driven shafts connected by a Dynaserv motor, relative motion between the outer channel and the central tool table can be achieved and controlled. The system has been designed such that the tool table can be stopped and raised out of the main testing area (known as parked ) while the channel continues to rotate. This allows the instrumented testing tools that are fixed to the actuator to be modified or changed without affecting the acceleration level on the soil. This was done between each test described here, allowing for the spudcans and caissons to be cleaned and interchanged. The primary advantage of using a drum centrifuge for the testing of shallow foundations is the large plan area of the test sample. The sample depth was approximately mm, with all tests performed at g; this corresponds to a prototype testing area of 3m by.5m and depth m. Further technical details of the drum centrifuge can be found in Stewart et al. [] and the various scaling relationships for modelling at enhanced accelerations are shown in Table. All results in this paper are presented in terms of prototype units. Loading Apparatus A loading leg that can be attached to the main radial actuator of the centrifuge has been designed and is shown in Figure 3. It is assembled from individual components allowing for a combination of tests. The central component can either be hinged or fully fixed, and the footing type can be altered. All footing diameters were kept constant at 6mm (prototype 6m) so that comparisons between footings could be carried out. The wall thickness for the caissons was mm. The spudcan and caisson attachments (skirt length to diameter ratio, L/D, of.5) are shown in Figure. By changing the skirt length, caissons with L/D ratios of.33 and. were also used. Photographs of the loading arm and footings are shown in Figure 5. The sign conventions for the loads and displacements are shown in Figure 6 and are based on those of Butterfield et al. [5]. Two load reference points were used when processing the caisson data: one at the base of the caisson skirts and one directly beneath the baseplate. For the spudcan the load reference point is the first point of maximum diameter, assuming the spudcan is penetrating into the soil. The load reference points are also indicated in Figure. Copyright by ASME

As indicated in Figure 3 the loading leg was strain-gauged in four locations and consisted of top and bottom axial gauges and top and bottom bending gauges. During the hinged arm test both sets of gauges could be used for the axial load whilst in the fixed arm tests only the bottom set of gauges could be used. The bending moments determined from the two sets of bending gauges could be used to determine the shear force. This shear force was then used to calculate the moment at the footing's load reference point using the appropriate lever arm. A laser was used to measure the distance between the fixed upper segment and the hinged lower section. The change in this distance could be translated into the angle of rotation of the hinge and hence the rotation of the footing. The components of the measured axial and shear forces were also resolved, based on the rotation, so that true vertical and horizontal loads acting on the footing could be determined. Vertical and horizontal displacements were obtained from the loading actuators. A specially designed valve was used during the caisson tests to allow for a realistic in-flight installation of the caisson [6]. This was attached to the baseplate and was left open until the caisson was installed (i.e. the soil plug reached the top of the inside section of the caisson). It could be closed by applying a pressure of kpa through an airline to the valve. Once the valve was closed a specific loading test could be carried out. The loading leg was designed with a hinge so that a combination of moment to horizontal load could be applied to the footing. The main test carried out during the combined loading investigation was a swipe test that gives an indication of a yield surface in combined load space [7, 8]. In order to carry out a test that is commensurate with the definition of a swipe test (i.e. with no vertical displacement of the reference point) an assumption is required about the movement of the hinge. In this instance a decision was taken to rotate the hinge in a circular arc about the load reference point for each footing. During the performance of the test there were small amounts of extra vertical, rotational and horizontal movements as the soil-structure system is not infinitely stiff. Of these movements the vertical is most critical with regards to the theoretical interpretation of the swipe test. The question is whether the surface tracked during the test approximates a yield surface. The initial hypothesis of the plasticity theories is that hardening of the yield surface is associated with vertical plastic penetration. In these tests the ratio of the elastic to plastic vertical stiffness is very large (of the order of ) so actually a small amount of vertical penetration would not cause any significant hardening of the yield surface. Therefore the shape of the surface that is swiped will give a close approximation to the actual yield surface. Soil Type The key properties of the Kaolin clay used are shown in Table. This clay has been used in many laboratory investigations at UWA and so has been well studied. The samples were prepared in a standard manner. The sample was approximately mm deep (m) with a normally consolidated profile. Soil characterisation tests were performed using a t-bar penetrometer [6]. These tests were performed at a rate of mm/s so that undrained behaviour was obtained. This can be assessed by examining the dimensionless velocity group, vd/c v, where v is the velocity, D is the appropriate length dimension and c v is the coefficient of consolidation. Finnie [9] suggests that if vd/c v is below. drained behaviour dominates whilst if vd/c v is above undrained behaviour dominates in this case vd/c v ~ 6 so undrained behaviour was obtained. One t-bar test was performed next to and on the same day as every footing test to get an estimate of the shear strength profile. A set of t-bar test results are shown in Figure 7 for test tank. They show a consistent strength profile that can be idealised as linearly increasing with a strength gradient of.5 kpa/m. EXPERIMENTAL RESULTS The experimental program consisted of two drum centrifuge samples. In each there was space for 3 individual footing test sites. The first sample concentrated on vertical compression/ tension behaviour, including the effect of cyclic loading at various stress levels. The second investigated combined loading with horizontal and moment loading applied during swipe, radial displacement and constant vertical load tests. Only selected results are presented here. The full collection of experimental results can be found in Cassidy and Byrne []. Vertical Loading (First Sample 3 tests) In the first sample 9 tests were used to investigate spudcan behaviour with the remainder investigating caisson behaviour. The test schedule for the vertical loading tests described in this paper is shown in Table 3. The results are illustrated in prototype units rather than non-dimensional quantities. Three caisson foundations were investigated with different skirt (L) to diameter (D) ratios. The most common L/D ratio was.5 but some tests were carried out with an L/D ratio of.33 and. This paper will discuss tests where the L/D ratio was.5 unless specifically mentioned. The first important vertical loading test is a monotonic push-in and pull-out, an example of which is illustrated in Figure 8 for a spudcan (BBMCT) and caisson (BBMCT5). The rate of testing was chosen so that the non-dimensional velocity group, vd/c v, was in the region of undrained behaviour. In this case the tests were carried out at a vertical displacement rate of. mm/s to give a vd/c v of ~9 (well above the undrained velocity). The spudcan load-displacement relationship shown in Figure 8 is a typical result. Initially as the spudcan embeds there is a rapid increase in strength until the full diameter is embedded in the soil. Upon full embedment the stiffness increases with penetration due to the increasing strength profile and the effect of the overburden. Note that in all spudcan tests there 3 Copyright by ASME

was significant backflow behind the penetrating spudcan. Conversely in the caisson test there is little resistance as the skirts are forced into the ground. The amount of this penetration resistance can be estimated using the idealised strength profile (.5 kpa/m). The appropriate expression for the effective vertical load (V ) is the sum of the internal and external frictional resistance plus the bearing capacity of the annulus of the skirt and is given by: V ( α D + α D ) s πh + ( N s γ ' h) πd t ' i i o o u c u + = where D is the diameter, α is an adhesion factor (typically taken as.5), h is the location of the base of the skirt from the mudline, t is the wall thickness, N c is a bearing capacity factor for an embedded strip footing (typically taken as 9), s u is the average shear strength over the depth of the skirt, s u is the shear strength at the base and γ is the effective unit weight (in this case 6kN/m 3 ). The subscripts i and o represent the inner and outer surfaces of the caisson. In the case when h = L (i.e. immediately after installation) the effective load required to push the caisson into the ground is approximately.mn which overpredicts slightly that obtained in the experiment (and shown in Figure 8). Naturally, once the baseplate makes contact with the ground there is a substantial increase in the vertical load. This response is very stiff compared to the spudcan response at a similar loading level. On further embedment of the caisson a distinct bearing capacity failure occurs. An estimate for this load can be obtained from: πd V ' = α odoπlsu + N c su where N c is a bearing capacity factor for the axisymmetric case. Initially this factor could be estimated by assuming the variation suggested by Skempton []: ( +.h / D ) 9 N c = 6 o 3 This variation was derived assuming a homogeneous clay deposit so a correction is required to account for the increasing soil strength with depth. Current practice (as reviewed by Martin [8]) for spudcan footings suggest taking the shear strength averaged over a depth of one radius below the footing. For the purpose of this paper it is assumed that this suggestion applies equally to caisson foundations. This leads to a calculated value for the bearing capacity of.mn, which compares well with the experimental value of MN. After yielding soft plastic behaviour occurs. The stiffness of the response of the caisson after yielding is similar to that of the spudcan. Furthermore, as can be seen in Figure 8 the initial stiffness and unload-reload stiffness for the caisson are of similar magnitudes to the unload-reload stiffness for the spudcan. The caisson was pushed in plastically for a further ave.67m to allow a comparison of this plastic behaviour with that of the spudcan. On the reversal of loading into tension there is substantial softening of the spudcan load-displacement response immediately on crossing the zero load axis. The caisson response differs in that there is no noticeable change in stiffness until well into the tensile region. If we consider the caisson to behave as a short stubby pile we can deduce that at small strains the tension is being taken as skirt friction. As the footing is pulled further from the ground then the bearing over the base will begin to dominate (i.e. at about % diameter). It is possible to estimate the amount of the tensile frictional resistance. This will be the sum of the external friction and the minimum of the internal friction and internal plug weight. For the case shown in Figure 8, where the caisson skirt tip is embedded to 5.67m, the result comes to. 9 MN. There is also likely to be a reverse bearing capacity failure around the annulus which would add another. 5 MN bringing the total tensile friction at small strains to. MN. This agrees well with the observed response as the load displacement curve only softens after a tensile load of. 5 MN. As the tensile failure progresses a strain level is reached where this initial frictional mechanism is superseded by a reverse bearing capacity failure acting over the base area. This load can be estimated similarly to Eq. : πd V ' = α odoπlsu + N csu Assuming that there is no degradation of the soil strength this would give a tensile failure load of. 95 MN. However, this lack of degradation is non-conservative as during the plastic penetration phase there is considerable remoulding of the soil taking place the caisson has penetrated approximately.5d. A reduction in shear strength to 7% of the original strength has been observed in repeated (remoulding) t-bar tests in kaolin clay [6]. If this reduction is adopted a reduced load of.35mn is obtained. This contrasts with the observed load of.95 MN in tension this would require a reduction in shear strength of 5%. It is possible to back calculate values of N c from the push in pull out tests performed. This allows comparisons to be made with published solutions. One appropriate equation (a rearrangement of equation ) to allow this back calculation is: N c V ' α o DoπLsu = 5 πd su This will require an estimate for the outside adhesion factor and in most cases might be taken as.5. Alternatively Martin and Randolph [] calculate an N c factor taking into account the effect of the skirt such that: Copyright by ASME

N c V ' = 6 πd su In their calculations they normalise to the shear strength at the skirt tip and account for the soil strength increasing with depth. Figure 9 shows the results calculated in this manner for the two tests discussed above so that they might be compared to solutions such as those provided by Martin and Randolph []. Figure shows the results of vertical loading tests for two caissons. The first test is that shown in Figure 8 (test BBMCT5) where the caisson is installed, then pushed a further.67m into the ground, before being withdrawn. In the second test (BBMCT6) the caisson was installed but then immediately withdrawn from the ground. During the installation the maximum compressive load was.6mn, well below the yield point. In contrast to BBMCT5 the tensile response exhibits a sudden yielding at a load of. MN. After reaching the peak there is considerable softening. Using Eq. the reverse bearing capacity is calculated as. MN, agreeing well with the experiment. Figure shows the effect of the skirt length on the pull out capacity of installed caissons (ones without compressive bearing capacity failure). The pull out response is shown for a L D =.5 caisson (BBMCT6) and a L D =. caisson (BBMCT5). Similar patterns of behaviour occur as immediately on being withdrawn a stiff response arises for both caissons. For the L D =. case the caisson was pushed back into the ground for m before being withdrawn again. A much softer response is again evident. The caisson was then pushed into the ground before being finally withdrawn. The peak loads after the plastic compressive loadings are much lower than the initial peak tensile load of. 3MN. This initial peak load compares with a theoretical value of. MN. Figure shows a more realistic loading regime for in-service caissons. This compares test BBMCT6, which is a pull-out test as described previously, with caissons that have undergone various amounts of cyclic loading. Test BBMCT7 has been installed then cycled about V = N. Test BBMCT8 was installed to a vertical load that was just past the compressive yield point, so there has been some plasticity occurring. The caisson is then cycled about V = N. There is no noticeable difference between the three tests except that the sharp peak of the monotonic pullout has been reduced to a more rounded peak for the tests where cyclic loading occurred. There is little accumulated displacement as the cyclic loading was two way about a mean of zero. Figure 3 shows a spudcan test (BBMCT9) and caisson test (BBMCT) where there has been cyclic loading around a mean load of ~5kN. This corresponds to V/V of.5 where V is the bearing capacity of the footing under pure vertical loading. The caisson has been installed just past the initial plastic yield point. Under the same loading regimes there is little difference in the response. In both cases there are accumulated displacements but as the mean load is compressive there is also some consolidation so the displacements are accumulating at a decreasing rate. This consolidation is also reflected in the higher compressive and tensile loads for both footings immediately after the cyclic loading. For instance the initial pullout of the spudcan, immediately after cycling, is. 9 MN compared to. 5 MN for test BBMCT. Figure shows the final set of results for the vertical loading. Test BBMCT8 has been described previously. This is compared to test BBMCT in which a spudcan footing has undergone the same loading pattern. The footing has been loaded to approximately MN before being withdrawn and cycled about a mean load of zero. The caisson shows no accumulated displacements. The spudcan on the other hand slowly pulls out of the ground with the cyclic loading. This is due to the asymmetric stiffness where any tension must be taken in the reverse bearing capacity mechanism rather than friction along the skirt sides. To mobilise the reverse bearing capacity mechanism requires some considerable strain level. The same strain level is not required to mobi lise a similar load in compression and so the footing pulls out of the ground. Combined Loading (Second Sample 3 tests) The majority of the combined loading tests performed were swipe tests where the footing is moved sideways or rotated whilst the vertical displacement is kept constant. It can be argued that the load path tracked is representative of a yield surface that could be used within a work hardening plasticity model [7, 8]. Figure 5 shows some typical results for a spudcan footing. The footing is penetrated into the ground. A swipe is performed after which the spudcan is penetrated further into the ground, erasing any memory of any previous event, so that another swipe test can be performed. In this way a large amount of information may be derived for particular footings in a minimum of tests. Usually the results in load space are normalised by V, which is the bearing capacity under pure vertical load, in the case of the spudcan it is the maximum value of the compressive load previously experienced by the footing. Figure 6 shows the results from Figure 5 normalised in this way. Figure 7a and 7b show fixed arm swipes of the spudcan, taken from a number of tests, and starting from different ratios of V/V (including negative values). These results are compared to other empirical data obtained by Martin [8]. For the results of these tests on kaolin clay Martin and Houlsby [3] have proposed the following empirical equation for the yield surface: 5 Copyright by ASME

M M + H H e M M H H β V V β β V V =...7 where M = m RV, H = hv, V V e = e + e, V V β ( β + β ) ( β ) + β = β β and β β V is the bearing capacity under pure vertical load. This surface was derived from a series of swipe experiments carried out on the laboratory floor [8]. The six parameter values that they obtained from regression analysis of the experimental data [3] were m =. 83, h =. 7, e =.58, e =. 8, β =. 76 and β =. 88. This surface is plotted on Figure 7a and 7b for the particular slice through the yield surface that is representative of the fixed arm swipes. It is clear that the surface provides a lower bound to the experimental data. Figure 8 shows the different ratios of M/RH for the two different loading arms indicating that the hinged arm allows the exploration of the {H:M/R} space not possible with a fixed arm. Figure 9 shows data for the hinged arm swipes for spudcans. In this case Martin s [8] surface provides a less convincing approximation of the experimental data in the horizontal plane. The moment plane is approximated closely by the above surface (equation 7). Finally Figure shows fixed arm swipes of both the spudcan and caisson starting from a V/V value of (i.e. the apex of the yield surface). The load reference point for the caisson has been taken at the base of the skirts so that it is comparable to the spudcan. In reality, considering structural analyses, the load reference point would be taken at the centre of the underside of the caisson baseplate, as this is reflective of where the caisson connects to the structure. Figure shows that the caisson load paths all exceed the spudcan loadpaths indicating that under similar conditions the caisson has a greater horizontal and moment capacity. Most of the tests for the caisson were carried out post yield point, however some were carried out prior to yield. One of these tests clearly has a larger horizontal capacity than the others. CONCLUSIONS This paper describes the background to a series of experiments performed on the drum centrifuge at UWA to assess the performance of spudcan and caisson foundations on normally consolidated soft clay. A number of vertical loading and combined loading tests were highlighted here, however, full details of the testing data is given in Cassidy and Byrne []. However, there are several concluding comments worth making. From the perspective of vertical loading there would appear to be much merit in replacing the spudcans with caisson foundations. Obviously there is a much greater resilience to tensile loading. Further analysis needs to be undertaken about the effect of cyclic loading and the amount of accumulated displacements that result. The caisson would be designed to be operating for its life within the very stiff pre-plasticity range. Though performed to make a comparative assessment of jackup foundations, the tests have also been used to help validate existing and develop new plasticity models. Integrating these into structural analysis packages (for jack-ups or even wind turbines), as well as experiments on calcareous soils, are the priorities for further research. ACKNOWLEDGMENTS This work was undertaken with support from Woodside Energy Limited (Po No:75) and an IREX grant from the Australian Research Council. The first author also acknowledges the support of the Royal Commission for the Exhibition of 85, the Apgar Prize and Magdalen College, Oxford. These experiments could not have been performed without the support of the drum centrifuge technician Mr Bart Thompson. The helpful advice and comments of Prof. Mark Randolph, Mr George Vlahos and Mr Andrew House were appreciated during the design of the loading leg and the course of the testing. The Centre for Offshore Foundation Systems was established and is funded under the Australian Government s Special Research Centres Program. REFERENCES [] Cassidy, M.J., Eatock Taylor, R., Houlsby, G.T. (). Analysis of jack-up units using a Constrained NewWave methodology. Applied Ocean Research 3, pp -3. [] Houlsby, G.T. and Byrne, B.W. (). Suction caisson foundations for offshore wind turbines and anemometer masts. Journal of Wind Engineering () pp 9-55. [3] Byrne, B.W. and Houlsby, G.T. (). Investigating novel foundations for offshore wind turbines. Proc. Int Conf. OMAE, Oslo, Norway. Paper N o OMAE-83. [] Stewart, D.P., Boyle, R.S. and Randolph, M.F. (998). Experience with a new drum centrifuge. Proc. Int. Conf. Centrifuge 98, Tokyo, Japan,, pp 35-. [5] Butterfield, R., Houlsby, G.T. and Gottardi, G. (997). Standardised sign conventions and notation for generally loaded foundations. Géotechnique 7 () pp 5-5. [6] Watson, P.G. (999). Performance of skirted foundations for offshore structures. PhD Thesis, University of Western Australia. [7] Tan, F.S.C. (99). Centrifuge and numerical modelling of conical footings on sand. PhD Thesis, University of Cambridge. [8] Martin, C.M. (99). Physical and numerical modelling of offshore foundations under combined loads. DPhil Thesis, University of Oxford. [9] Finnie, I.M.S. (993). Performance of shallow foundations in calcareous soils. PhD Thesis, University of Western Australia, Perth. [] Cassidy, M.J. and Byrne, B.W. (). Drum centrifuge model 6 Copyright by ASME

tests comparing the performance of spudcans and caissons in kaolin clay. OUEL Report N o 8/, Department of Engineering Science, The University of Oxford. [] Skempton, A.W. (95). The bearing capacity of clays. Proc Building Research Congress, London,, pp 9-89. [] Martin, C.M. and Randolph, M.F. (). Applications of the lower and upper bound theorems of plasticity to collapse of circular foundations. Report N o G57, Geomechanics Group, UWA. [3] Martin, C.M. and Houlsby, G.T. (). Combined loading of spudcan foundations on clay: laboratory tests. Géotechnique 5 () pp 35-338. [] Stewart, D.P. (99). Lateral loading of piled bridge abutments due to embankment construction. PhD Thesis, UWA. TABLES Table Scaling relationships used during centrifuge modelling. Quantity Relationship (model/prototype) Gravity N Stress Strain Length /N Force /N Moment /N 3 Density /N Mass /N 3 Time (consolidation) /N Table Kaolin clay properties []. Property Value Liquid Limit, LL 6 % Plastic Limit, PL 7 % Plasticity Index, I P 3 % Specific Gravity, G s.6 Angle of Internal Friction, φ 3 Consolidation Coefficient (mean), c v m /year Submerged unit weight, γ 6.8 kn/m 3 Table 3 Vertical loading tests referred to in this paper. Test Name Footing Comments BBMCT Spudcan Vertical unload-reload @.mm/s BBMCT Spudcan N N Cycle @ Vm=N (±N) pull-out BBMCT5 L/D=.5 Vertical unload-reload @.mm/s pull-out BBMCT6 L/D=.5 Pull-out after installation 5N pull-out BBMCT7 L/D=.5 Installation 5N N 5 Cycles @ V m =N (±N, N, 5N) pull-out BBMCT8 L/D=.5 Installation N N 5 Cycles @ V m =N (±N, N, 5N) pull-out BBMCT9 Spudcan N 5N Cycle @ V m =5 (±N, N, 3N, 5N, 6N) pull-out BBMCT L/D=.5 Installation N 5N 5 Cycles @ V m =5N (±N) pull-out BBMCT5 L/D=. Vertical unload-reload @.mm/s pull-out 7 Copyright by ASME

Anemometer mast or turbine support structure (a) (b) (c) Water surface Seabed Caissons Caisson Steel pile NOT TO SCALE Figure - Options for offshore wind applications (after Houlsby and Byrne []). Figure - The UWA drum centrifuge. Figure 3 - Design of loading leg used during experiments. 8 mm thread 8 mm thread "poppet" valve Load Reference Point Spudcan Caisson L/D=.5 Load Reference Points Figure - Model spudcan and caisson footings. 8 Copyright by ASME

Figure 5a - Photograph of spudcan and fixed loading arm. Figure 5b - Photograph of caisson and hinged loading arm. R Reference position w M H Current position u V θ Figure 6 - Sign convention after Butterfield et al. [5]. Undrained Shear Strength, s u (kpa) 6 8 3 5 6 7 8 9 Figure 7 - Profiles of undrained shear strength as deduced from t-bar tests. 9 Copyright by ASME

BBMCT BBMCT5 BBMCT BBMCT5.5.5 -.5 - Bearing Capacity Factor, N c 8 6 - - -6-8 -.5 3 5 6 Figure 8 - Loading of caisson and spudcan. - 3 5 6 Figure 9 - Determining the bearing capacity factor. BBMCT5 BBMCT6 BBMCT6 BBMCT5.5 3.5 -.5 - - - -.5 3 5 6 Figure - Caisson loading and pullout. -3 6 8 Figure - Caisson pullouts. Copyright by ASME

BBMCT6 BBMCT7 BBMCT8 BBMCT9 BBMCT.5 3.5.5 -.5.5.5 -.5 - - -.5 -.5 3 Figure - Pullout of caissons. - 6 8 Figure 3 - Cyclic loading of footings..5 BBMCT BBMCT8. - -.5.5.5.5 -.5 - -.5 3 Figure - Further cyclic loading. Horizontal Load, H (MN).. 3.. 5. 6. 7..3.5..5..5. - -.5.5.5 Figure 5 - Swipe tests. Copyright by ASME

H /V o M /RV o.8.6....8.6... -..6.5..3... -. -....6.8....6.8. Figure 6 - Normalised swipe tests. H /V o.5..5..5 -.5 -.5.5.5 Martin's Equation [8] Figure 7a - Fixed arm swipes of spudcan..7..6.8 Hinged Arm.5. Martin's Equation [8].6 M /RV o.3. M /RV o. Fixed Arm.. -. -. -.5.5.5 Figure 7b - Fixed arm swipes of spudcan. -. -.5.5..5..5 H /V o Figure 8 - Loadpaths in H /V o :M /RV o space. Copyright by ASME

.7..6.5 Martin's Equation [8].8..6 H /V o.3. M /RV o. Martin's Equation [8].. -. -. -.3 -.5.5.5 -. -.5.5.5 Figure 9 - Hinged arm swipes of spudcan. BBMCT7 BBMCT8 Yield Surface BBMCT7 BBMCT8 Yield Surface.5..5..8.35.6 H /V o.3.5..5 Martin's Equation [8] M /RV o.. Martin's Equation [8]..5 -. -.5...6.8. -....6.8. Figure - Fixed arm swipes of caisson (BBMCT7) and spudcan (BBMCT8). 3 Copyright by ASME